The server is saying that it is receiving acks from a node that it didn’t expect to have sent a particular type of request to. Since the other node seems to know the in progress transaction-id on the complaining node suggests that the complaining node actually sent the request to the acking node.
You have also observed this issue, Cluster Visibility error, which hints that there may be a cluster configuration issue. Do these nodes have multiple NICs (logical of physical)? If so, could you run ip addr across all node to verify that no two nodes report the same address? You could also try configuring the access-address directly.
Ah, docker networking can be tricky. The simplest solution would be to use host networking.
In your configuration, are x.x.x.x or y.y.y.y publicly accessible IP addresses? If they are, you should reconsider running a backed service on a public IP - if not, is really necessary to obfuscate them?
In your config, all x.x.x.x except access-address should be 172.17.0.1 (or any). The y.y.y.y should be the IP of the remote host. Also if the MAC addresses of the 172.17.0.1 interfaces are the same on each container then you will also need to override the node-id.